CN108734178A - A kind of HOG feature extracting methods of rule-basedization template - Google Patents
A kind of HOG feature extracting methods of rule-basedization template Download PDFInfo
- Publication number
- CN108734178A CN108734178A CN201810477502.1A CN201810477502A CN108734178A CN 108734178 A CN108734178 A CN 108734178A CN 201810477502 A CN201810477502 A CN 201810477502A CN 108734178 A CN108734178 A CN 108734178A
- Authority
- CN
- China
- Prior art keywords
- block
- hog feature
- image detection
- hog
- extraction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of HOG feature extracting methods of rule-basedization template, by the correlation that the step-length of Cell sizes, Block sizes and each Block movements is arranged, determine that HOG feature extraction regularization templates carry out the extraction of HOG features, the personnel that can direct study well carry out the extraction and calculating of HOG characteristic values, the time that researcher adjusts model is saved, and makes final results contrast accurate.
Description
Technical field
The invention belongs to computer visions and technical field of image processing, and in particular to a kind of rule-basedization template
The design of HOG feature extracting methods.
Background technology
Histograms of oriented gradients (Histogram of Oriented Gradient, HOG) is characterized in one kind in computer
It is used for carrying out the Feature Descriptor of object detection in vision and image procossing.In image recognition especially in pedestrian detection, HOG is special
Sign and SVM classifier broad incorporation, obtain great success.
During extracting HOG features, there are many parameters that researcher is needed to go to design according to the specific requirement of task
And debugging, but these work are completed without a set of rule that can be instructed or follow.Because parameter is more, obtain compared with
Good model just needs researcher to take a lot of time debugging and modification parameter, and due to no highly effective tune
Examination, amending method, finally obtained result may be also undesirable.
Invention content
The purpose of the present invention is to solve existing HOG feature extracting methods, to take long and result not necessarily accurate
The problem of, it is proposed that a kind of HOG feature extracting methods of rule-basedization template.
The technical scheme is that:A kind of HOG feature extracting methods of rule-basedization template, include the following steps:
S1, the size that image detection block is chosen according to the complexity of image detection task, obtain image detection block.
S2, pixel redistribution processing is carried out to image detection block.
To the pixel value equal proportion scaling of image detection block so that pixel average 128.
S3, using Edge extraction operator (such as Roberts operators, Sobel operators, Prewitt operators, Canny operators
Or Log operators) extraction image detection block edge, obtain image border figure.
S4, HOG feature extraction regularization templates are determined according to the size of image detection block.
Regularization is carried out to the step-length of Cell sizes, Block sizes and each Block movement of HOG feature extractions, is obtained
To HOG feature extraction regularization templates, it is specifically configured to:
Cell is dimensioned to by the height and width of image detection block to be divided exactly, and it is big that Block is dimensioned to 3*3 Cell
Small, the step-length of each Block movement is set as the integral multiple of Cell sizes, and (such as is set as one less than Block sizes
Cell sizes).
S5, image border figure is divided using HOG feature extraction regularization templates, and extracts HOG features.
The beneficial effects of the invention are as follows:The HOG feature extracting methods of rule-basedization template provided by the invention can be very
The personnel that direct study well carry out the extraction and calculating of HOG characteristic values, save researcher and adjust the time of model, and make
The results contrast obtained finally is accurate.
Description of the drawings
Fig. 1 show a kind of HOG feature extracting method flow charts of rule-basedization template provided in an embodiment of the present invention.
Specific implementation mode
Carry out detailed description of the present invention illustrative embodiments with reference to the drawings.It should be appreciated that shown in attached drawing and
The embodiment of description is only exemplary, it is intended that is illustrated the principle and spirit of the invention, and is not limited the model of the present invention
It encloses.
An embodiment of the present invention provides a kind of HOG feature extracting methods of rule-basedization template, as shown in Figure 1, including
Following steps:
S1, the size that image detection block is chosen according to the complexity of image detection task, i.e. H*W (high * wide) obtain image
Detection block.
Image in the embodiment of the present invention is gray-scale map, and usual image detection task is more complicated, and object deformation is more, image
Detection block is bigger.Such as:When detecting pedestrian, 56*20 sizes may be selected in image detection block representative value;When detecting vehicle, image inspection
It surveys block representative value and 25*30 sizes may be selected.If in addition, training dataset is sufficiently large, and precision prescribed is higher, can choose larger
Image detection block.
S2, pixel redistribution processing is carried out to image detection block.
Due to extraction image gradient or edge when, to strength sensitive, thus need to image detection block carry out pixel
Redistribution processing.In the embodiment of the present invention, the specific method that pixel redistribution processing is carried out to image detection block is:Image is examined
Survey the pixel value equal proportion scaling of block so that pixel average 128, gradient and edge are more obvious.
S3, using Edge extraction operator (such as Roberts operators, Sobel operators, Prewitt operators, Canny operators
Or Log operators) extraction image detection block edge, obtain image border figure.Using above-mentioned each Edge extraction operator extraction
The routine techniques that the specific method of image border is known in the art, details are not described herein.
S4, HOG feature extraction regularization templates are determined according to the size of image detection block.
Regularization is carried out to the step-length of Cell sizes, Block sizes and each Block movement of HOG feature extractions, is obtained
To HOG feature extraction regularization templates, it is specifically configured to:
(1) Cell sizes:First, the limited size of Cell is in the image detection block size of setting, the size of Cell (it is high and
Width is respectively how many pixel) it wants by the height and width of image detection block to be divided exactly;Also, it after determining Cell, is made of Cell
The number of pixels that Block includes cannot be very few (very few that excessively sparse -0 yuan of characteristic value can be caused excessive).
(2) Block sizes:Common Block is dimensioned to 3*3 Cell size, if selection larger size, unit
Characteristic value in Block can compacter (0 yuan tails off), but final characteristic parameter dimension can become smaller, and be unfavorable for more complicated
Classification task.
(3) step-length that Block is moved every time:Block moving step lengths are set as the integral multiple of Cell sizes, but generally can be small
A Cell size is set as in the size of Block, the embodiment of the present invention.But when the number of Cell is excessive, Block
Moving step length is too small to cause characteristic parameter dimension excessive, influence to detect real-time, at this moment can suitably increase step-length, experiment card
It is bright detection accuracy is influenced in the case of this it is little.
S5, image border figure is divided using HOG feature extraction regularization templates, and extracts HOG features.
Those of ordinary skill in the art will understand that the embodiments described herein, which is to help reader, understands this hair
Bright principle, it should be understood that protection scope of the present invention is not limited to such specific embodiments and embodiments.This field
Those of ordinary skill can make according to the technical disclosures disclosed by the invention various does not depart from the other each of essence of the invention
The specific variations and combinations of kind, these variations and combinations are still within the scope of the present invention.
Claims (5)
1. a kind of HOG feature extracting methods of rule-basedization template, which is characterized in that include the following steps:
S1, the size that image detection block is chosen according to the complexity of image detection task, obtain image detection block;
S2, pixel redistribution processing is carried out to image detection block;
S3, using the edge of Edge extraction operator extraction image detection block, obtain image border figure;
S4, HOG feature extraction regularization templates are determined according to the size of image detection block;
S5, image border figure is divided using HOG feature extraction regularization templates, and extracts HOG features.
2. HOG feature extracting methods according to claim 1, which is characterized in that the step S2 is specially:Image is examined
Survey the pixel value equal proportion scaling of block so that pixel average 128.
3. HOG feature extracting methods according to claim 1, which is characterized in that the image border in the step S3 carries
It is Roberts operators, Sobel operators, Prewitt operators, Canny operators or Log operators to take operator.
4. HOG feature extracting methods according to claim 1, which is characterized in that the step S4 is specially:To HOG spies
The step-length for levying Cell sizes, Block sizes and each Block movement of extraction carries out regularization, obtains HOG feature extractions rule
Then change template, is specifically configured to:
Cell is dimensioned to by the height and width of image detection block to be divided exactly, and Block is dimensioned to 3*3 Cell size,
The step-length of each Block movements is set as the integral multiple of Cell sizes, and is less than Block sizes.
5. HOG feature extracting methods according to claim 4, which is characterized in that the step-length of each Block movements is set
It is set to a Cell size.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810477502.1A CN108734178A (en) | 2018-05-18 | 2018-05-18 | A kind of HOG feature extracting methods of rule-basedization template |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810477502.1A CN108734178A (en) | 2018-05-18 | 2018-05-18 | A kind of HOG feature extracting methods of rule-basedization template |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108734178A true CN108734178A (en) | 2018-11-02 |
Family
ID=63937609
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810477502.1A Pending CN108734178A (en) | 2018-05-18 | 2018-05-18 | A kind of HOG feature extracting methods of rule-basedization template |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108734178A (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102622584A (en) * | 2012-03-02 | 2012-08-01 | 成都三泰电子实业股份有限公司 | Method for detecting mask faces in video monitor |
CN103440478A (en) * | 2013-08-27 | 2013-12-11 | 电子科技大学 | Face detection method based on HOG characteristics |
CN104778453A (en) * | 2015-04-02 | 2015-07-15 | 杭州电子科技大学 | Night pedestrian detection method based on statistical features of infrared pedestrian brightness |
CN107862680A (en) * | 2017-10-31 | 2018-03-30 | 西安电子科技大学 | A kind of target following optimization method based on correlation filter |
CN108038846A (en) * | 2017-12-04 | 2018-05-15 | 国网山东省电力公司电力科学研究院 | Transmission line equipment image defect detection method and system based on multilayer convolutional neural networks |
-
2018
- 2018-05-18 CN CN201810477502.1A patent/CN108734178A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102622584A (en) * | 2012-03-02 | 2012-08-01 | 成都三泰电子实业股份有限公司 | Method for detecting mask faces in video monitor |
CN103440478A (en) * | 2013-08-27 | 2013-12-11 | 电子科技大学 | Face detection method based on HOG characteristics |
CN104778453A (en) * | 2015-04-02 | 2015-07-15 | 杭州电子科技大学 | Night pedestrian detection method based on statistical features of infrared pedestrian brightness |
CN107862680A (en) * | 2017-10-31 | 2018-03-30 | 西安电子科技大学 | A kind of target following optimization method based on correlation filter |
CN108038846A (en) * | 2017-12-04 | 2018-05-15 | 国网山东省电力公司电力科学研究院 | Transmission line equipment image defect detection method and system based on multilayer convolutional neural networks |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109886928B (en) | Target cell marking method, device, storage medium and terminal equipment | |
WO2019096174A1 (en) | Method and apparatus for garment color identification of figure images, and electronic device | |
US20160364849A1 (en) | Defect detection method for display panel based on histogram of oriented gradient | |
CN106651872A (en) | Prewitt operator-based pavement crack recognition method and system | |
CN110930390B (en) | Chip pin missing detection method based on semi-supervised deep learning | |
CN110210448B (en) | Intelligent face skin aging degree identification and evaluation method | |
US20090226047A1 (en) | Apparatus and Method of Processing Image and Human Face Detection System using the smae | |
KR102073468B1 (en) | System and method for scoring color candidate poses against a color image in a vision system | |
CN110781877B (en) | Image recognition method, device and storage medium | |
CN110807775A (en) | Traditional Chinese medicine tongue image segmentation device and method based on artificial intelligence and storage medium | |
CN104835175A (en) | Visual attention mechanism-based method for detecting target in nuclear environment | |
US10192283B2 (en) | System and method for determining clutter in an acquired image | |
WO2017120796A1 (en) | Pavement distress detection method and apparatus, and electronic device | |
CN108108753A (en) | A kind of recognition methods of check box selection state based on support vector machines and device | |
Shaikh et al. | A novel approach for automatic number plate recognition | |
CN108664970A (en) | A kind of fast target detection method, electronic equipment, storage medium and system | |
CN106228157A (en) | Coloured image word paragraph segmentation based on image recognition technology and recognition methods | |
CN112991374A (en) | Canny algorithm-based edge enhancement method, device, equipment and storage medium | |
CN110599453A (en) | Panel defect detection method and device based on image fusion and equipment terminal | |
CN103699876B (en) | Method and device for identifying vehicle number based on linear array CCD (Charge Coupled Device) images | |
CN108009574A (en) | A kind of rail clip detection method | |
US11068740B2 (en) | Particle boundary identification | |
CN107578001B (en) | Method and device for testing resolution of fingerprint acquisition equipment | |
Ashourian et al. | Real time implementation of a license plate location recognition system based on adaptive morphology | |
CN116883893A (en) | Tunnel face underground water intelligent identification method and system based on infrared thermal imaging |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181102 |
|
RJ01 | Rejection of invention patent application after publication |